Executive summary

This research addressed two questions:

  1. Are standardized butterfly traps an effective way to quantify the butterfly fauna of the Mohonk Preserve?
  2. Has climate change impacted emergence time of the butterfly fauna of the Mohonk Preserve?

    1. During the late Spring and early Summer of 2016 I established transects for standardized butterfly traps at three locations on the Mohonk Preserve: Spring Farm, White Oak, and Glory Hill. These traps were baited with fermented bananas and monitored following a protocol established by Conservation International for fruit-feeding butterflies. This was the first application of this protocol in North America. Briefly, traps were baited and checked daily for three days and any butterflies present in the traps were identified, marked, and released. Other butterflies encountered in proximity to the traps were also identified and their abundances noted. Following each survey period, the traps were idle for two weeks and the trapping cycle resumed. After three trapping periods (nine survey days over six weeks) a total of three butterfly species were captured in the traps with no recaptures. I observed an additional 12 species of butterfly in significant abundances at the sites that were never observed in the traps. The low butterfly diversity observed in the traps, and higher diversity outside the traps, strongly suggests that this method is not ideal to monitor butterfly biodiversity at the Mohonk Preserve.

    2. The Mohonk Preserve has collected weather data continuously for over 100 years and observed the first occurrence of butterfly species for nearly as long. I was able to access the preserve’s weather data for dates beginning in 1896. Using a variety of statistical methods I detected the signature of climate change at the Mohonk Preserve. Since 1980 the average annual temperature has increased by 1\(^{\circ}\)C at the Mohonk Preserve. Using statistical modelling, I predict that this upward trend is likely to continue with a 0.25\(^{\circ}\)C increase over the next 10 years. I requested but was not given the corresponding butterfly observational data and could not investigate how this temperature increase has affected the butterfly fauna of the Mohonk Preserve.

    All data and the R statistical code required to reproduce this document are freely available on the internet and can be found at this projects GitHub page. Please switch to the gh-pages branch to ensure that you acquire the correct code and data used to render this report.


Introduction

Species Diversity

Standardized trap sampling fruit-feeding nymphalid butterflies has been shown to be an effective means for understanding butterfly diversity in space and time, and for use in conservation efforts (DeVries and Walla 2001; Hill and Hamer 2004; Molleman et al. 2006; DeVries et al. 2012; Freitas et al. 2014). Conservation International, a world leader in assessing biodiversity, has produced a standard protocol to estimate butterfly biodiversity that can be employed anywhere in the world (DeVries et al. 2016).

The data generated by these trap studies will allow the estimation of population size as well as species richness turnover, which are important first steps in monitoring the health and viability of local species. Once determined, the Mohonk Preserve will be able to monitor these populations (perhaps incorporating citizen scientists) and detect changes in the preserve before they are visually evident (Agosti et al. 2000).

The Nymphalidae is the largest family of butterflies, and the fruit-feeding butterflies may comprise up to 50% of the nymphalid species richness in tropical forests (DeVries et al. 2012).

Fruit-feeding nymphalids at the Mohonk preserve include members of the genera Nymphalis, Speyeria, Phyciodes, Vanessa, and Polygonia. One of the most salient characteristics of this group is that they have been sampled in a standardized manner to avoid human collector biases, thus facilitating comparisons of species richness, composition and abundance within and among habitat types. For these reasons, I propose focusing standardized sampling methods exclusively on fruit-feeding nymphalids, rather than on the entire butterfly community. There are many trap studies now being conducted all over the world.

Most of these are, however, not directly comparable because they do not use consistent trap designs and protocols (see examples and citations in DeVries 1987, DeVries and Walla 2001, Batra (2006), Freitas et al. (2014)). The sampling protocol provided here is based on more than 10 years of monthly sampling conducted in lowland forests at Garza Cocha, Sucumbios Province, Ecuador and the Tirimbina Biological Reserve Heredia Province, Costa Rica that have been demonstrated to be directly comparable (DeVries and Walla 2001, DeVries et al. (2012)).

Global climate change

Global climate change is one of the most serious threats to biodiversity of natural and human ecosystems, affecting both the distribution of species and the timing of biological events (Amano et al. 2010). Since the 1980’s mean annual temperature in the Northeastern United States has risen approximately 2\(^{\circ}\)F. Some of this increase has occurred during the late winter and early spring seasons, resulting in a seasonal warm up 5-10 days earlier than was the case in 1980 (Andresen and Winkler 2009). Herbivorous insects, such as butterflies, must be synchronized with their host plants such that food resources are available to the insects at critical stages of development. If this symbiosis is disturbed, as has been predicted with global change, there may be limited resources available for developing larvae (Parmesan 2007).

Following an approach similar to that of Cook et al. (2008), I planned use the wealth of natural history observations available at the Mohonk Preserve, such as first occurrence records, to correlate emergence with existing weather data from the Mohonk Preserve. In addition to the data collected by Mohonk Preserve personnel, the American Museum of Natural History has numerous specimens from the area in their collection, many dating back over 100 years. The relationship between insect emergence and accumulated degree days (a measure of total heating) is well established and remains constant for particular species (Pedigo 2002). Significant warming has occurred in the Northeastern United States since 2002 (IPCC 2013). If warming has occurred at the Mohonk Preserve the calendar date of an event should occur earlier in the year and is detectable (Barnett et al. 1999). I then planned to use projections of future climate change to examine how the butterfly community may respond to the predicted changes (Winkler et al. 2011).

Data on the first observation of butterfly species were not provided. As such, I was unable to perform the phenology analysis. I chose to address and related and important question: is climate change occurring at the Mohonk Preserve? To address this question, I used the Mohonk Preserve temperature data set and applied statistical forecasting models.


Trap study

Locations

Transects were established at Spring Farm, White Oak, and Glory Hill.

**Figure 1**: Trap locations on Mohonk Preserve

Figure 1: Trap locations on Mohonk Preserve

Modified from DeVries et al. (2016): A completed trap is a cylinder 1 m tall and 37 cm in diameter with a closed top and open bottom (Fig 2). Two metal ring frames are sewn into the top and bottom, and the netting must completely close the top of the cylinder. The cylinder was sewn such that the netting overlaps on the long axis by 2 cm leaving a 20 cm unsewn slit approximately 30 cm from the top to allow access to the trap interior. A 47-49 cm square trap base (3 mm of durable corrugated plastic) was suspended from the bottom ring of the cylinder such that is hung 6 cm below the opening of the cylinder. The diameter of the trap base extended 5-6 cm beyond the cylinder diameter. Holes were drilled on each side, and plastic cable ties were used to attach the base to the trap.

A small plastic bait cup was secured to the center of the base with a loop of thin, stiff wire that passed through two holes drilled in the base. The wire was then pressed down into the mouth of the cup to keep the bait cup upright and centered on the base. The receptacle for the bait had a volume of at least 200 ml (8 ounces), and was just be tall enough to pass between the base and lower trap ring. A length of nylon cord was secured to the bottom of the trap base to assist securing traps.

Bait: Traps were baited with locally obtained bananas that were first chopped into 2-3 cm pieces and pureed with an auger in a large container (that had a lid). Approximately 15 pounds of bananas were used over the course of this project. The pureed bananas were allowed to ferment in the sealed container for 48 hours prior to use. The day before trapping approximately 150-200 ml of banana mash was added to the bait receptacle in each trap such that the bait level is below the top of the receptacle. Sampling began on the next day. Following each sampling period the banana puree was removed from the trap.

**Figure 2**: Trap construction diagram

Figure 2: Trap construction diagram

Diversity

Observed diversity

Below are the cumulative counts of butterflies recorded over three trapping periods. During this period, the individual traps were accessable to butterflies a total of 3240 hours (9 days, 3 site, 5 traps / site, 24 hours / day).

Species Count
Cercyonis pegala 6
Megisto cymela 14
Papilio polyxenes 6

Table 1: Species encountered in traps and the sum of all counts.

Species Observed
Battus philenor 2
Celastrina ladon 2
Cercyonis pegala 5
Chlosyne harrisii 1
Coenonympha tullia 6
Colias eurytheme 2
Cupido comyntas 4
Danaus plexippus 3
Enodia anthedon 3
Epargyreus clarus 3
Euphydryas phaeton 3
Erynnis juvenalis 5
Junonia coenia 2
Limenitis arthemis 1
Megisto cymela 4
Papilio canadensis 1
Papilio cresphontes 1
Papilio glaucus 2
Papilio polyxenes 1
Phyciodes tharos 4
Pieris virginiensis 4
Polygonia comma 1
Nymphalis antiopa 6
Nymphalis vaualbum 1
Satyrodes eurydice 4
Speyeria cybele 2
Vanessa virginiensis 3

Table 2: Species observed in close proximity to traps.

All of the butterfliy species observed in proximity to the traps were cross-checked to records available on the e-butterfly website or in the collection of the American Museum of Natural History. All of the butterflies I report above have been observed in county.

Estimated diversity

Using the data generated from the trap survey we can calculate “numbers equivalents” for a variety of indices. A numbers equivalent is the effective number of species for a given index at a particular order of diversity measure (q). q is a weight that is applied to rare things; when q > 1 the analysis is more sensitive to common species (rare things have little impact on the estimate. ) When q < 1 the analysis is more sensitive to rare species. When estimating values of \(\gamma\) at q = 0 the result is the species richness.

**Figure 3**: Species richness estimates based on traping.

Figure 3: Species richness estimates based on traping.

Butterfly sampling using baited traps at the Mohonk Preserve did not produce species richness estimates that were indicative of the actual species present in the habitats (Tables 1 & 2). This research did analytically derived the correct species richness values for within the traps, but these estimates (or 3 species for Spring Farm and White Oak, 2 species for Glory Hill) were far below the observed values found immediatley outside of the trap. This was the first application of this method in North America, and has at least demonstrated how not to estimate butterfly biodiversty in New York. Future attempts to assess the butterfly community at the Mohonk Preserve should utilize alternative methods.


Climate change

Introduction

Climate change relates to an extended change in the distribution of temperatures that an area experiences (IPCC 2013). Climate change has been observed throughout the United States and is particularly evident in the Nort

(Knutson et al. 2017)

When evaluating data, an important (and often overlooked) step is to visualize the data set in an informative manner. The data set provided by the Mohonk Preserve contains 42,499 entries since 1900-01-01 (January \(1^{st}\), 1900). Generating a coherant plot that contains over 42,000 data points would be problematic to say the least. To increase the interpretability of the data I will utilize some form of summarized data, either mean monthly or annual temperatures. For example, it may be informative to consider how temperature may have changed by month.

Temperature by month

A straight forward way to visual temperature trends is to plot data by some shared trait. For example, plotting the montly average temperature for the month of December across all years. Plotting data in this manner removes seasonal variation and allows one to explore linear trends.

**Figure 4**: Mean temperature by month and year.

Figure 4: Mean temperature by month and year.

By removing the seasonality found in the Mohonk Preserve temperature data we can apply simple linear models to the each month. It appears that most months, especially March - August, have experienced visible warming trends, wheras

While this figure provides a useful graphic that is easy to interpret, we can fit more complex models to the data set in order to forcast future temperatures at the Mohonk Preserve. I will ask two primary questions of the data:

  1. Is climate change detectable using the temperature data that the Mohonk Preserve has collects? I will address this question using a very flexible generalized additive mixxed model (gamm).
  2. Do forecasting models predict future warming for the Mohonk Preserve? I will employ two different modeling approaches to address this question. The AutoRegressive Integrated Moving Average model (ARIMA) and a Bayesian variant of the gamm each make different assumtions about the underlying nature of the data. If both models agree we can have higher confidence in the results.

General additive mixed model

We seek a model that explains the underlying data. Given the complex and fluctuating nature of temperature data it makes little sense to apply a linear model . I followed the method of Morice et al. (2012) and applied a generalized additive mixed model (gamm) that fits a series of local regressions to the data (thin-plate regression spline) that allows incorporation of the correlated residuals. This analysis was implemented using the gamm() function of the mgcv package (Wood 2011, Wood (2013), Wood (2006)). In order to account for the seasonal fluctuations that are found in continuous temperature data we will incorporate a smoothing parameter.

**Figure 5**: Plot of mean annual temperature with `gamm` line.

Figure 5: Plot of mean annual temperature with gamm line.

The gamm was run using “centered” temperature data to ease interpretability, this standardization has no effect on the analysis. When data are centered, the aritmatic mean is subtracted from each value, resulting in an average temperature of 0\(^{\circ}\)C.

This model indicates that the the average annual temperature of the Mohonk Preserve has increased ~1.2\(^{\circ}\)C since 1975. This result is consistent with many previous studies from the northeastern United States that demonstrated an annual increase in temperature since the early 1980’s (Andresen and Winkler 2009, Knutson et al. (2017)).

Prediction - ARIMA

I applied an ARIMA model to forecast the future temperature changes at the Mohonk preserve. ARIMA models are especially good at predicting future values in time-series data (Hyndman and Athanasopoulos 2013).

(Hyndman and Khandakar 2008, Hyndman (2017))

**Figure 6**: `ARIMA` model with 10 year forecast

Figure 6: ARIMA model with 10 year forecast

Forecast is the mean of a series of simulated futures based on the model that best fits our data.

The forecast predicts that annual temperature will continue to be above the mean for the past 115 years.

Prediction - prophet

**Figure 7**: `prophet` model with 10 year forecast

Figure 7: prophet model with 10 year forecast

Summary

Cook et al. (2008) missed it, the trend was evident in 2002.


Literature cited

Agosti, D., J. Majer, L. Alonso, and T. Schultz. 2000. Ants: standard methods for measuring and monitoring biodiversity. Smithsonian Institution Press.

Amano, T., R. J. Smithers, T. H. Sparks, and W. J. Sutherland. 2010. A 250-year index of first flowering dates and its response to temperature changes. Proceedings Of The Royal Society B-Biological Sciences 277:2451–2457.

Andresen, J. A., and J. A. Winkler. 2009. Weather and climate of michigan. in D. B. R.J. Schaetzl J.T. Darden, ed. Michigan geography and geology. Pearson Custom Publishing, Boston, MA.

Barnett, T. P., K. Hasselmann, M. Chelliah, T. Delworth, G. Hegerl, P. Jones, E. Rasmusson, E. Roeckner, C. Ropelewski, B. Santer, and S. Tett. 1999. Detection and attribution of recent climate change: A status report. Bulletin of the American Meteorological Society 80:2631–2659.

Batra, P. 2006. Butterfly monitoring protocol. in Tropical Ecology, Assessment, and Monitoring (TEAM) Initiative. Conservation International.

Cook, B. I., E. R. Cook, P. C. Huth, J. E. Thompson, A. Forster, and D. Smiley. 2008. A cross-taxa phenological dataset from Mohonk Lake, NY and its relationship to climate. International Journal of Climatology 28:1369–1383.

DeVries, P. J., L. G. Alexander, I. A. Chacon, and J. A. Fordyce. 2012. Similarity and difference among rainforest fruit-feeding butterfly communities in Central and South America. Journal of Animal Ecology 81:472–482.

DeVries, P. J., and T. R. Walla. 2001. Species diversity and community structure in neotropical fruit-feeding butterflies. Biological Journal of the Linnean Society 74:1–15.

DeVries, P. J., C. A. Hamm, and J. A. Fordyce. 2016. A standardized sampling protocol for fuit-feeding butterflies (Nymphalidae). Pp. 139–148 in T. H. Larsen, ed. Core standardized methods for rapid biological field assessment. Conservation International.

Freitas, A., C. A. Iserhard, J. P. Santos, J. Y. Carreira, D. B. Ribeiro, D. H. Alves, A. H. Rosa, O. J. Marini-Filho, G. M. Accacio, and M. Uehara-Prado. 2014. Studies with butterfly bait traps: an overview. Revista Colombiana de Entomologia 40:209–218.

Hill, J. K., and K. C. Hamer. 2004. Determining impacts of habitat modification on diversity of tropical forest fauna: the importance of spatial scale. Journal of Applied Ecology 41:744–754.

Hyndman, R. J. 2017. Forecast: Forecasting functions for time series and linear models.

Hyndman, R. J., and G. Athanasopoulos. 2013. Forecasting: Principles and practice. OTexts.

Hyndman, R. J., and Y. Khandakar. 2008. Automatic time series forecasting: the forecast package for R. Journal of Statistical Software 26:1–22.

IPCC. 2013. Climate change 2013: The physical science basis. contribution of working group I to the fifth assessment report of the intergovernmental panel on climate change. Cambridge University Press.

Knutson, T., J. P. Kossin, C. Mears, J. Perlwitz, and M. F. Wehner. 2017. Detection and attribution of climate change. Pp. 114–132 in K. A. H. D J Wuebbles D W Fahey, ed. Climate Science Special Report: Fourth National Climate Assessment, Volume I. U.S. Global Change Research Program.

Molleman, F., A. Kop, P. M. Brakefield, P. J. De vries, and B. J. Zwaan. 2006. Vertical and Temporal Patterns of Biodiversity of Fruit-Feeding Butterflies in a Tropical Forest in Uganda. Biodiversity and Conservation 15:107–121.

Morice, C. P., J. J. Kennedy, N. A. Rayner, and P. D. Jones. 2012. Quantifying uncertainties in global and regional temperature change using an ensemble of observational estimates: The HadCRUT4 data set. Journal of Geophysical Research: Atmospheres 117:2156–2202.

Pedigo, L. P. 2002. Entomology and Pest Management. Prentice Hall.

Winkler, J. A., G. S. Guentchev, M. Liszewska, and P. Tan. 2011. Climate Scenario Development and Applications for Local/Regional Climate Change Impact Assessments: An Overview for the Non-Climate Scientist. Geography Compass 5:310–328.

Wood, S. N. 2011. Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. Journal of the Royal Statistical Society, Series B (Statistical Methodology) 73:3–36.

Wood, S. N. 2006. Generalized additive models: an introduction with R. Chapman Hall/CRC.

Wood, S. N. 2013. On p-values for smooth components of an extended generalized additive model. Biometrika 100:221–228.